Technology outsmarts the human investor, MARCH 8, 2017 by: John Gapper
To grasp why Standard Life agreed to buy Aberdeen Asset Management for £3.8bn this week, it helps to recall an experiment at the Oregon Research Institute in the 1960s, that rated how good doctors were at making judgments compared with formulas they helped to design. The models won, as Michael Lewis records in The Undoing Project, his book about the economics Nobel Prize-winning psychologist Daniel Kahneman and his late collaborator Amos Tversky. Not only did it cost less to use them to identify cancers and psychological disorders in patients than doctors, but they were more accurate. The Oregon study found that it was best to check the results of clinical tests with an algorithm once professionals had made the rules for diagnosis. In contrast to technology, humans got tired and distracted, and had off days. “Only rarely — if at all — will the [outcomes] favour the continued employment of a man over a model of a man,” concluded a pivotal research paper in 1970. That judgment hangs over the men and women of the investment management industry as well as the medical profession. So-called alpha — the inimitable talent of skilled human investors — is worth paying for, as is treatment by expert doctors. But formulas are perfectly good at doing the predictable stuff, and often better. It took time for technology to be able to match what people do in asset management but the era has arrived. It is evident at Aberdeen, which has suffered a £105bn net outflow of funds since 2013 as investors have turned away from human expertise, as well as its speciality of emerging markets. Aberdeen’s problem is common to many traditional asset managers and hedge funds since the 2008 crisis, and is prompting cost-saving mergers such as the Standard Life deal. As interest rates and investment returns have fallen, investors are less willing to pay for human interference. Too many active managers have charged a lot for little more than matching the returns investors receive from automated passive funds; on average, active managers lag indices such as the S&P 500 when fees and costs are accounted for. Money is steadily flowing into index and exchange traded funds, which Moody’s estimates may overtake active funds in the US by 2024. Much of this was foreseen. In 1961, when scientific and other professions were more male-dominated than today, two researchers concluded: “Men and computers could co-operate more efficiently . . . if a man could tell the computers how he wanted decisions made, and then let the computers make the decisions for him.” It is an uncannily accurate description of what is known as “smart beta” — the freedom to pick aspects of investment return, such as risk, growth and volatility, and replicate them in automated form through an exchange traded fund. Once instructed how to select from thousands of shares or bonds, algorithms can ensure that any fund has the right blend of attributes. Technology has made inroads into what money managers do. First came funds that matched indices so that an investor could achieve the same return as, for example, the FTSE 100 without having to buy all 100 securities and keep on updating them. Then came smart beta’s ability to reproduce professional investment styles equally precisely and more cheaply. “It just gets harder and harder and harder,” reflected one money manager this week. His is the predicament of other professionals — anything done by a person that follows a pattern and can be coded into a form that a computer understands will soon get squeezed. Technology also has the advantage identified in 1970: algorithms stay constantly alert. It does not imply the complete death — or automation — of the investment manager. A professional can still undertake original research on a company or a security that provides insight. As more of the market becomes automated, originality becomes rarer and more valuable: an idiosyncratic investor should achieve higher returns by standing out from the robotic crowd. Nor can algorithmic efficiency be wholly divorced from human intelligence, as the Oregon study showed — the point was that humans needed to set parameters for computers to follow. Many asset managers use analysts and researchers to build investment models that then trade securities automatically; others blend their active risk-taking with passive elements. This could be a low moment for human asset managers — not only do their fees look especially high in a low interest rate world but quantitative easing inflates asset prices and makes it harder to be distinctive. There are cyclical reasons for the hard times. But these difficulties demonstrate how automation eats into professions, not by taking away all the jobs in one day but by unbundling them — dividing them between tasks that only humans can perform and those of which an algorithm is quite capable. Then the boundary relentlessly shifts. The Oregon researchers showed clearly half a century ago what would happen, and now it has.